Good afternoon,
I am looking at these GPUs, I was wondering what you guys think? And if anyone knows the RAC for the 1080ti.

---

2070S > 1080ti > 1080

with current WU distribution, a 1080ti should land around 100-110k RAC (on the CUDA Special application) maybe 1/3 of that on the SoG app

---

Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

2070S > 1080ti > 1080

with current WU distribution, a 1080ti should land around 100-110k RAC (on the CUDA Special application) maybe 1/3 of that on the SoG app

you think the super 2070 is the best to get?

---

yes definitely. the 2070 Super will be noticeably faster than the 1080ti.

---

Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

2070S > 1080ti > 1080

with current WU distribution, a 1080ti should land around 100-110k RAC (on the CUDA Special application) maybe 1/3 of that on the SoG app

I am running 3 2070's my RAC is in the 60-70k range. How can 1-1080 produce 110-110?

---

I was talking the 1080ti, not 1080.

But the answer is in that I’m talking about different applications. The 1080ti is slightly slower than a 2070. And on the CUDA special app, each 2070 will do about 110-115k RAC. Reference my 10x 2070 system which is doing 1.13 Million right now. The special app is Linux only. You can’t run it on Windows.

By contrast you’re using the SoG app. And to be honest I’d think your system would be closer to 100k. Either my memory that the the Special App is 3x faster than SoG is off, or maybe you have something else going on with your system (thermal issues, not running 24/7, suspend while computer in use, etc)

---

Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

2070S > 1080ti > 1080

with current WU distribution, a 1080ti should land around 100-110k RAC (on the CUDA Special application) maybe 1/3 of that on the SoG app

I am running 3 2070's my RAC is in the 60-70k range. How can 1-1080 produce 110-110?

@Carlos
If you are staying with Windows It looks like you are already running the Lunatics Distro.

Download/Install "GPU-Z" and take a look at how loaded your gpu's are. If they are not running at 95% mostly then there is an issue. If there are a lot of "saw tooth's" rather than running pretty steadily then there is something reducing your production.

It looks like you are already running command line for your gpus. You can try running

-period_iterations_num 1

(check my spelling!) and see if it makes any difference.

You are already running -tt 1500 I think. Not sure what size -sbs you are running but I certainly would try 2048 rather than something smaller.

Your app_config.xml should have 1 cpu per 1 gpu thread.

It is possible you can get more total production if you run 2 gpu tasks / card. Only testing will confirm that.

Your cpu should be running 6 or 7 cores/threads out of 8 so that the cpu is not over committed. You can set % of cores in the Boinc Manager (local mode) or on the Website.

I hope these ideas are new to you. If they are not, I apologize.

Tom

---

A proud member of the OFA (Old Farts Association).

I seem to be doing all that, but sill only seeing daily credit in the 70k range. As for GPU I am seeing 100% with some fluctuation, but not much. I think it's time for a CPU update, which is on order.
Thanks for looking Tom.

What are the temps and clock speeds of the cards? If they are thermal throttling, they will run slower.

Also running 2 WU at a time will boost performance a little. The 2070 should be able to handle it. But I would reduce the sbs value to 1024. I think I remember someone saying that values over 1024 don’t do anything and it will just default back to 1024 in practice.

---

Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

No you can run a single task with a -sbs 2048 value. I did it. In the stderr.txt output, you will see something like a 2136MB buffer being used. But that only works for singles because of the OpenCL 26% VRAM limit which ends up around ~2MB on the typical 8MB VRAM card. If you have a 1080 Ti or Titan with 11GB or 12GB then you can use a 3076MB buffer for a single task.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)